north of 30°N over the period 1900–2005 but downward trends dominate the tropics since the 1970s. AR4 included analysis of both the GHCN (Vose et al., 1992) and CRU (Mitchell and Jones, 2005) gaugebased precipitation data sets for the globally averaged annual precipitation over land. For both data sets the overall linear trend from 1900 to 2005 (1901–2002 for CRU) was positive but not statistically significant (Table 3.4 from AR4). Other periods covered in AR4 (1951– 2005 and 1979–2005) showed a mix of negative and positive trends depending on the data set.
Since AR4, existing data sets have been updated and a new data set developed. Figure 2.28 shows the century-scale variations and trends on globally and zonally averaged annual precipitation using five data sets: GHCN V2 (updated through 2011; Vose et al., 1992), Global Precipitation Climatology Project V2.2 (GPCP) combined raingauge–satellite product (Adler et al., 2003), CRU TS 3.10.01 (updated from Mitchell and Jones, 2005), Global Precipitation Climatology Centre V6 (GPCC) data set (Becker et al., 2013) and a reconstructed data set by Smith et al. (2012). Each data product incorporates a different number of station series for each region. The Smith et al. product is a statistical reconstruction using Empirical Orthogonal Functions, similar to the NCDC MLOST global temperature product (Section 2.4.3) that does provide coverage for most of the global surface area although only land is included here. The data sets based on in situ observations only start in 1901, but the Smith et al. data set ends in 2008, while the other three data sets contain data until at least 2010.
For the longest common period of record (1901–2008) all datasets exhibit increases in globally averaged precipitation, with three of the four showing statistically significant changes (Table 2.9). However, there is a factor of almost three spread in the magnitude of the change which serves to create low confidence. Global trends for the shorter period (1951–2008) show a mix of statistically non-significant positive and negative trends amongst the four data sets with the infilled Smith et al. (2012) analysis showing increases and the remainder decreases. These differences among data sets indicate that long-term increases in global precipitation discussed in AR4 are uncertain, owing in part to issues in data coverage in the early part of the 20th century (Wan et al., 2013).
In summary, confidence in precipitation change averaged over global land areas is low for the years prior to 1950 and medium afterwards because of insufficient data, particularly in the earlier part of the record. Available globally incomplete records show mixed and non-significant long-term trends in reported global mean changes. Further, when virtually all the land area is filled in using a reconstruction method, the resulting time series shows less change in land-based precipitation since 1900.